Evaluating the Effects of Treatment Regimes in the Presence of Drop-Outs: Application to Depression Data

• Activity Number: 461
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Yenchih Hsu*+ and Adbus S. Wahed
• Companies: University of Pittsburgh and University of Pittsburgh
• Keywords: Dynamic treatment regime ; Generalized estimating equations ; Inverse-probability-weighting ; Missing data ; Two-stage designs

Abstract

Studies in mental health research often employ two-stage designs to assess combinations of pharmacotherapies and psychotherapies. Outcomes in such studies often consist of repeated measurements of scores such as the 24-item Hamilton Rating Scale for Depression (HRSD) over a fixed duration of time. Since treatment is given sequentially, the eligibility of receiving one treatment assignment depends on previous treatments and intermediate outcomes. The goal is to compare different treatment regimes in two-stage longitudinal studies to find the most beneficial one for each patient. The presence of missing data is a common phenomenon in longitudinal studies (drop-outs, withdrawals, etc.). In this paper, we show how to construct proper weights to account for monotone missing and apply them in the estimating equations to draw inference for treatment regimes from two-stage longitudinal studies. Specifically, we provide consistent estimators and their asymptotic variances of the effects of treatment regimes. Large-sample properties of the proposed estimators are provided analytically, and examined through simulations. We apply our methods to a depression dataset that motivated this study.